Middle Data Scientist (Matching)

TL;DR

Middle Data Scientist (Matching): Leading ML projects end-to-end, from problem definition to production and support, with an accent on developing and training models for matching, ranking, and nearest neighbor search. Focus on ensuring model quality after launch, including metrics, monitoring, drift/degradation analysis, and improvement plans.

Локация: Россия, удалённо

Salary: 280 000 — 350 000 ₽/мес

Компания

Островок is a travel-tech company creating online booking platforms for hotels, air tickets, and other services for individual travelers, corporate clients, and travel agencies.

Что делать

• Lead ML projects end-to-end: problem definition -> solution -> test -> production -> support.
• Collaborate with data engineers to form datasets and data requirements, assess feasibility, risks, and limitations.
• Participate in the design and analysis of A/B tests with data analysts: metrics, splits, interpretation of results, recommendations for rolling out solutions.
• Develop and train models (classic ML + DL), including solutions on text and image embeddings; conduct offline evaluation and error analysis.
• Transfer the model and code to production (Python service), support releases and integrations.
• Be responsible for the quality of the model after launch: metrics, monitoring, drift/degradation, improvement plan, and support regulations.

Требования

• 2+ years of experience as a Data Scientist (in matching, ranking, nearest neighbor search, etc.).
• Experience leading ML projects end-to-end in production (from setup to support).
• Excellent understanding of classic ML: feature engineering, boosting, classification/regression, cross-validation, threshold selection, calibration.
• Experience with DL (PyTorch/TensorFlow): understanding the principles of fine-tuning, model inference.
• Python (production quality): readable code, tests for critical components, understanding of model/artifact packaging and integration into the service.
• Understanding of ML monitoring: quality metrics, drift, alerts, diagnostics, and support regulations.
• SQL at the level of independent dataset assembly (joins, window functions).
• Experience with interpretability and error analysis of the model.
• MLflow / W&B / DVC or similar experiment tracking tools.
• Orchestration/pipelines (Airflow/Prefect/Dagster) and advanced data processes.
• Разговорный английский на уровне В1.

Культура и преимущества

• Interesting projects: creating products for travelers, travel agents, and hoteliers around the world.
• Full freedom to achieve results: flexible schedule, remote or office — you decide where and when to work.
• Non-standard approach to work and a thirst for new things, for example, we solve some problems with AI.
• Technical community Ostrovok! Tech holds meetups, hackathons, participates in conferences, and supports even the most daring ideas.
• Professional development: helping employees speak at conferences — from submitting an application to preparing a presentation.
• Taking care of the team's well-being: from the first month of work, our employees have VMI and discounts in the Yasno service.
• Internal adaptation and training programs, soft skills and leadership development, tailored individually for each employee.
• Partial compensation for participation in external training and conferences.
• Learning English: corporate group classes, conversation clubs, and discounts on Skyeng courses.
• Corporate prices for hotels and other travel services — so that our employees travel more often.
• Ostrovok is an accredited IT company.